A Prediction Based Long-Cycle Time Synchronization Algorithm for Sensor Networks

Existing time synchronization algorithms and protocols mostly focus on improving the synchronization accuracy. However, they usually require frequent resynchronization to keep designed precision in actual applications, which leads to high energy consumption and heavy traffic load. This paper presents a Prediction based Long-cycle Time Synchronization algorithm (PLTS), which puts emphasis on reducing the resynchronization frequency while guaranteeing a given accuracy. PLTS is a combination of periodic synchronization and prediction synchronization. It makes use of an existing time synchronization protocol to accomplish the periodic synchronization, while during the intervals of periodic synchronization, each node applies a prediction model to calibrate its own logic time according to the crystal oscillator´s frequency characteristics. By this means, all nodes can keep synchronization till next periodic synchronization starts. Experiment results show that PLTS can reduce resynchronization frequency remarkably and possesses good merits in saving energy and reducing traffic load.